Data visualization with ggplot2

Day 2 - Introduction to Data Analysis with R

Selina Baldauf

Freie Universität Berlin - Theoretical Ecology

September 20, 2024

A ggplot showcase

Example plots you can create with ggplot

A ggplot showcase

Visualization by Jake Kaupp, code available on Github

A ggplot showcase

Visualizations produced by the BBC News data team

A ggplot showcase

Visualization by Cédric Scherer, code available on Github

Advantages of ggplot

  • Consistent grammar/structure
  • Flexible structure allows you to produce any type of plots
  • Highly customizable appearance (themes)
  • Many extension packages that provide additional plot types, themes, colors, animation, …
    • See here for a list of ggplot extensions
  • Active community that provides help and inspiration
  • Perfect package for exploratory data analysis and beautiful plots

The data

Data set and_vertebrates with measurements of a trout and 2 salamander species in different forest sections.

  • year: observation year
  • section: CC (clear cut forest) or OG (old growth forest)
  • unittype: channel classification (C = Cascade, P = Pool, …)
  • species: Species measured
  • length_1_mm: body length [mm]
  • weight_g: body weight [g]

Coastal giant salamander (terrestrial form)
Andrews Forest Program by Lina DiGregorio via CC-BY from https://andrewsforest.oregonstate.edu

The data

Data set and_vertebrates with measurements of a trout and 2 salamander species in different forest sections.

library(lterdatasampler)
and_vertebrates
#> # A tibble: 32,191 × 6
#>     year section unittype species         length_1_mm weight_g
#>    <dbl> <chr>   <chr>    <chr>                 <dbl>    <dbl>
#>  1  1987 CC      R        Cutthroat trout          58     1.75
#>  2  1987 CC      R        Cutthroat trout          61     1.95
#>  3  1987 CC      R        Cutthroat trout          89     5.6 
#>  4  1987 CC      R        Cutthroat trout          58     2.15
#>  5  1987 CC      R        Cutthroat trout          93     6.9 
#>  6  1987 CC      R        Cutthroat trout          86     5.9 
#>  7  1987 CC      R        Cutthroat trout         107    10.5 
#>  8  1987 CC      R        Cutthroat trout         131    20.6 
#>  9  1987 CC      R        Cutthroat trout         103     9.55
#> 10  1987 CC      R        Cutthroat trout         117    13   
#> # ℹ 32,181 more rows

Artwork by Allison Horst

ggplot(data)

The ggplot() function initializes a ggplot object. Every ggplot needs this function.

library(ggplot2)
# or library(tidyverse)

ggplot(data = and_vertebrates)
  • Empty plot because we did not specify the mapping of data variables

aes(x, y)

The aesthetics define how data variables are mapped plot properties.

ggplot(
  data = and_vertebrates,
  mapping = aes(
    x = length_1_mm,
    y = weight_g
  )
)
  • Map variable length_1_mm to x-axis and weight_g to y-axis
  • Default scales are automatically adapted to range of data

aes(x, y)

The aesthetics define how data variables are mapped plot properties.

ggplot(
  data = and_vertebrates,
  mapping = aes(
    x = length_1_mm,
    y = weight_g
  )
)

This is the same but shorter:

ggplot(
  and_vertebrates,
  aes(
    x = length_1_mm, 
    y = weight_g)
)

Remember argument matching by position?

geom_*

geoms define how data points are represented. There are many different geoms to chose from

from ggplot cheatsheet

geom_point

ggplot(
  data = and_vertebrates,
  aes(
    x = length_1_mm,
    y = weight_g
  )
) +
  geom_point()
  • New plot layers added with +
  • Warning that points could not be plotted due to missing values
  • data and aes defined in ggplot call are inherited to all plot layers
#> Warning: Removed 13270 rows containing missing values (`geom_point()`).

geom_point

ggplot() +
  geom_point(
    data = and_vertebrates,
    aes(
      x = length_1_mm,
      y = weight_g
    )
  )
  • data and aes can also be local to a layer:

Here, it does not make a difference in the result.

#> Warning: Removed 13270 rows containing missing values (`geom_point()`).

aes(color): mapping color to a variable

Looks like there are two groups of data: Map color of points to a variable by adding it to aesthetics:

ggplot(
  data = and_vertebrates,
  aes(
    x = length_1_mm,
    y = weight_g,
    color = species
  )
) +
  geom_point()
  • Map the species variable to the color aesthetic of the plot

aes(size): mapping size to a variable

We can do the same with size:

ggplot(
  data = and_vertebrates,
  aes(
    x = length_1_mm,
    y = weight_g,
    size = species
  )
) +
  geom_point()

aes(shape): mapping shape to a variable

We can do the same with shape:

ggplot(
  data = and_vertebrates,
  aes(
    x = length_1_mm,
    y = weight_g,
    shape = species
  )
) +
  geom_point()

Combine color, size and shape

We can also combine these aesthetics and map different variables

ggplot(
  data = and_vertebrates,
  aes(
    x = length_1_mm,
    y = weight_g,
    color = unittype,
    shape = species,
    size = year
  )
) +
  geom_point()
  • This is a bit too much for this plot, but sometimes can be useful

Changing the scales of the aesthetics

The scales onto which the aesthetic elements are mapped can be changed.

ggplot(
  data = and_vertebrates,
  aes(
    x = length_1_mm,
    y = weight_g,
    color = species
  )
) +
  geom_point()
  • Exponential relationship?
  • How does it look like on the log scale?

scale_x_log10

The scales onto which the aesthetic elements are mapped can be changed.

ggplot(
  data = and_vertebrates,
  aes(
    x = length_1_mm,
    y = weight_g,
    color = species
  )
) +
  geom_point() +
  scale_x_log10() +
  scale_y_log10()
  • Scales can be changed for all elements of aes:

scale_aes-name_scale-type

In this example we scale the x and the y aesthetic to log10.

geom_smooth

Add a smoothing line that helps see patterns in the data

ggplot(
  data = and_vertebrates,
  aes(
    x = length_1_mm,
    y = weight_g,
    color = species
  )
) +
  geom_point(alpha = 0.3) +
  geom_smooth(method = "lm") +
  scale_x_log10() +
  scale_y_log10()
  • With method = "lm", a linear regression line is added
  • All geoms done separately for species because color is defined globally
  • Alpha makes points transparent (0-1)

geom_boxplot

Compare groups using a boxplot

ggplot(
  and_vertebrates,
  aes(
    x = species,
    y = length_1_mm
  )
) +
  geom_boxplot()

geom_boxplot

Compare groups using a boxplot

ggplot(
  and_vertebrates,
  aes(
    x = species,
    y = length_1_mm
  )
) +
  geom_boxplot(notch = TRUE)
  • If notches don’t overlap, the medians of the groups are likely different

geom_boxplot

Map the unittype to the color aesthetic of the boxplot

ggplot(
  and_vertebrates,
  aes(
    x = species,
    y = length_1_mm,
    color = unittype
  )
) +
  geom_boxplot()

geom_boxplot

Map the unittype to the fill aesthetic of the box

ggplot(
  and_vertebrates,
  aes(
    x = species,
    y = length_1_mm,
    fill = unittype
  )
) +
  geom_boxplot(notch = TRUE)

geom_histogram

ggplot(
  and_vertebrates,
  aes(
    x = length_1_mm,
    fill = section
  )
) +
  geom_histogram()
  • Careful: By default the histogram is stacked for the different groups!

geom_histogram

ggplot(
  and_vertebrates,
  aes(
    x = length_1_mm,
    fill = section
  )
) +
  geom_histogram(
    position = "identity",
    alpha = 0.5
  )
  • Change the position of the histogram to "identity", if you don’t want it stacked
  • alpha makes sure that you see overlapping areas

geom_tile

You can create a simple heatmap with geom_tile

ggplot(
  and_vertebrates,
  aes(
    x = section,
    y = species,
    fill = weight_g
  )
) +
  geom_tile()
  • Here we would have to choose a different color scheme to see differences

Small multiples with facet_wrap

Split your plots along one variable with facet_wrap

ggplot(
  data = and_vertebrates,
  aes(
    x = length_1_mm,
    y = weight_g,
    color = species
  )
) +
  geom_point() +
  facet_wrap(~section)

Small multiples with facet_grid

Split your plots along two variables with facet_grid

ggplot(
  data = and_vertebrates,
  aes(
    x = length_1_mm,
    y = weight_g,
    color = unittype
  )
) +
  geom_point() +
  facet_grid(section ~ species)
  • facet_grid(rows ~ columns)

Now you

Task 1.1 - 1.2 (45 min)

Exploratory data analysis with the penguin data set

Find the task description here

(Don’t do the “Beautify” task)

Artwork by Allison Horst

Change appearance of points

ggplot(and_vertebrates, aes(
  x = length_1_mm,
  y = weight_g
)) +
  geom_point(
    size = 4,
    shape = 17,
    color = "blue",
    alpha = 0.5
  )

Shape and color codes

Change color scale

We can also save a plot in a variable

g <- ggplot(and_vertebrates, aes(
  x = length_1_mm,
  y = weight_g,
  color = species
)) +
  geom_point()

g


  • Other plot layers can still be added to g

scale_color_viridis_d

Change the colors of the color aesthetic:

g +
  scale_color_viridis_d(
    option = "cividis"
  )


  • The viridis color palette is designed for viewers with common forms of color blindness
  • Different options of viridis color palettes: "magma", "inferno", "plasma", "viridis", "cividis"

scale_color_manual

We can also manually specify colors:

g +
  scale_color_manual(
    values = c(
      "darkolivegreen4",
      "darkorchid3"
    )
  )
  • Length of color vector has to match number of levels in your aesthetic
  • Specify colors
    • Via their name
    • Via their Hex color codes (use websites to generate your own color palettes, e.g. here)

Other color scales

You can use the paletteer package to access color scales from many packages.

# install.packages("paletteer")
library(paletteer)
g <- g +
  scale_color_paletteer_d(
    palette = "ggsci::default_uchicago"
  )
g
  • Use scale_color_paletteer_d for discrete and scale_color_paletteer_c for continuous color scales
  • Check out all palettes available here

scale_fill_* vs. scale_color_*

ggplot(
  and_vertebrates,
  aes(
    x = section,
    y = length_1_mm,
    color = unittype)) +
  geom_boxplot() +
  scale_color_paletteer_d(
    palette = "ggsci::default_uchicago"
  )

ggplot(
  and_vertebrates,
  aes(
    x = section,
    y = length_1_mm,
    fill = unittype)) +
  geom_boxplot() +
  scale_fill_paletteer_d(
    palette = "ggsci::default_uchicago"
  )

labs: Change axis and legend titles and add plot title

g <- g +
  labs(
    x = "Length [mm]",
    y = "Weight [g]",
    color = "Species",
    title = "Length-Weight relationship",
    subtitle = "There seems to be an exponential relationship",
    caption = "Data from the `lterdatasampler` package"
  )
g

labs: Change axis and legend titles and add plot title

theme_*: change appearance

ggplot2 offers many pre-defined themes that we can apply to change the appearance of a plot.

g +
  theme_classic()

g +
  theme_bw()

theme_*: change appearance

ggplot2 offers many pre-defined themes that we can apply to change the appearance of a plot.

g +
  theme_minimal()

g +
  theme_dark()

theme(): customize theme

You can manually change a theme or even create an entire theme yourself. The elements you can control in the theme are:

  • titles (plot, axis, legend, …)
  • labels
  • background
  • borders
  • grid lines
  • legends

If you want a full list of what you can customize, have a look at

?theme
  • Look here for an overview of the elements that you can change and the corresponding functions

theme(): customize theme

To edit a theme, just add another theme() layer to your plot.

g +
  theme_minimal() +
  theme(
    legend.position = "bottom",
    axis.text = element_text(
      face = "bold"
    ),
    plot.background = element_rect(
      fill = "lightgrey",
      color = "darkred"
    )
  )

The basic functioning of theme elements is:

theme(
  element_name = element_function()
)

theme_set(): set global theme

You can set a global theme that will be applied to all ggplot objects in the current R session.

# Globally set theme_minimal as the default theme
theme_set(theme_minimal())

Add this to the beginning of your script.

You can also specify some defaults, e.g. the text size:

theme_set(theme_minimal(base_size = 16))

This is very practical if you want to achieve a consistent look, e.g. for a scientific journal.

ggsave()

A ggplot object can be saved on disk in different formats.

Without specifications:

# save plot g in img as my_plot.pdf
ggsave(filename = "img/my_plot.pdf", plot = g)
# save plot g in img as my_plot.png
ggsave(filename = "img/my_plot.png", plot = g)

Or with specifications:

# save a plot named g in the img directory under the name my_plot.png
# with width 16 cm and height 9 cm
ggsave(
  filename = "img/my_plot.png",
  plot = g,
  width = 16,
  heigth = 9,
  units = "cm"
)

Have a look at ?ggsave to see all options.

Now you

Task 2 (30 min)

Make your penguin plots more beautiful

Find the task description here

Selina Baldauf // Data visualization with ggplot2